Religious Portrait Thangka Image Retrieval Based on Gesture Feature

摘要

This paper proposed a novel method for religious portrait Thangka image retrieval based on gesture feature. Firstly, adopt the method of TPLR (two point locate rule) which we propose to annotate gesture of images that will be saved to image database, extract gesture features which are represented in a quintuple vector to store in characteristic library; Secondly, the same task such as gesture annotation, feature extraction, in the process of image retrieval to be treated to the retrieved image, then compute similarity between sample image and the image of image database using the algorithm this paper presented; Finally, we apply the method of ITFA (interval threshold filter algorithm) to obtain the final retrieval result. The experiments show that the proposed algorithm can retrieve the similar images well and efficient.